diff --git a/.github/ci-setup-action/action.yml b/.github/ci-setup-action/action.yml
index 15df4df1a28..0c6fee9bbb5 100644
--- a/.github/ci-setup-action/action.yml
+++ b/.github/ci-setup-action/action.yml
@@ -18,7 +18,7 @@ runs:
       id: cache-submodules
       uses: actions/cache@v3
       with:
-        path: .git
+        path: .git/modules
         key: submodules-${{ hashFiles('.gitmodules') }}
 
     - name: Checkout Submodules
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index d42849a778a..e1415d124e7 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -6,9 +6,34 @@ on:
   workflow_dispatch: {}
 
 jobs:
+  start-runner:
+    timeout-minutes: 5              # normally it only takes 1-2 minutes
+    name: Start self-hosted EC2 runner
+    runs-on: ubuntu-latest
+    permissions:
+      actions: write        
+    steps:      
+      - name: Start EC2 runner
+        id: start-ec2-runner
+        uses: AztecProtocol/ec2-action-builder@v11
+        with:
+          github_token: ${{ secrets.GH_SELF_HOSTED_RUNNER_TOKEN }}
+          aws_access_key_id: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          aws_secret_access_key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+          aws_region: "us-east-2"
+          ec2_instance_type: r6a.32xlarge
+          ec2_ami_id: ami-04d8422a9ba4de80f
+          ec2_subnet_id: subnet-4cfabd25
+          ec2_instance_tags: '[]'
+          github_action_runner_version: v2.315.0
+          ec2_security_group_id: sg-0ccd4e5df0dcca0c9
+          ec2_instance_ttl: 40 # 40 minutes to reap
+          ec2_spot_instance_strategy: BestEffort
+
   # there's a lot of x86 tasks - let's split out the build step
   build-x86:
-    runs-on: ubuntu-latest
+    needs: start-runner
+    runs-on: ${{ github.run_id }}
     outputs:
       e2e_list: ${{ steps.e2e_list.outputs.list }}
     env:
@@ -17,6 +42,8 @@ jobs:
     concurrency:
       group: build-${{ github.ref_name == 'master' && github.run_id || github.ref_name }}-x86
       cancel-in-progress: true
+    # permission check
+    if: github.event.pull_request.head.repo.full_name == github.repository || contains(github.event.pull_request.labels.*.name, 'external-ci')
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -45,7 +72,8 @@ jobs:
 
   # all the end-to-end integration tests for aztec
   e2e-arm:
-    runs-on: ubuntu-latest
+    needs: start-runner
+    runs-on: ${{ github.run_id }}     
     env:
       EARTHLY_TOKEN: ${{ secrets.EARTHLY_TOKEN }}
     strategy:
@@ -54,9 +82,8 @@ jobs:
         test:
           - e2e-card-game
           - e2e-crowdfunding-and-claim
-    # cancel if reran on same PR if exists, otherwise if on same commit
     concurrency:
-      group: ${{ matrix.test }}-${{ github.ref_name == 'master' && github.run_id|| github.ref_name }}-arm
+      group: ${{ matrix.test }}-${{ github.ref_name == 'master' && github.run_id || github.ref_name }}-arm
       cancel-in-progress: true
     steps:
       - name: Checkout
@@ -82,7 +109,7 @@ jobs:
   # all the end-to-end integration tests for aztec
   e2e-x86:
     needs: build-x86
-    runs-on: ubuntu-latest
+    runs-on: ${{ github.run_id }}     
     env:
       EARTHLY_TOKEN: ${{ secrets.EARTHLY_TOKEN }}
     strategy:
@@ -120,9 +147,14 @@ jobs:
           # we blank earthly token just to be sure this is a pure local run
           EARTHLY_TOKEN="" earthly -P --no-output +${{ matrix.test }} --e2e_mode=cache
 
+      - name: Upload logs
+        run: |
+          BRANCH=${{ github.ref_name }} PULL_REQUEST=${{ github.event.number }} scripts/ci/upload_logs_to_s3 ./yarn-project/end-to-end/log
+
   # barretenberg (prover) native tests
   bb-native-tests:
-    runs-on: ubuntu-latest
+    needs: start-runner
+    runs-on: ${{ github.run_id }}     
     env:
       EARTHLY_TOKEN: ${{ secrets.EARTHLY_TOKEN }}
     strategy:
@@ -157,8 +189,9 @@ jobs:
   # they should use parallelism within the benchmark, but only one thing should run at a time
   # for accurate results
   # We don't depend on 'build' as we use a different runner and will build components on the fist step that uses them.
-  bench:
-    runs-on: ubuntu-latest
+  bb-bench:
+    runs-on: ${{ github.run_id }}
+    needs: start-runner
     env:
       EARTHLY_TOKEN: ${{ secrets.EARTHLY_TOKEN }}
     # cancel if reran on same PR if exists, otherwise if on same commit
@@ -200,3 +233,20 @@ jobs:
       - name: Ultrahonk Bench
         working-directory: ./barretenberg/cpp/
         run: earthly-cloud bench x86 --no-output +bench-ultra-honk --bench_mode=cache
+
+  # # Post actions, deploy and summarize logs
+  # aztec-bench-summary:
+  #   runs-on: ${{ github.run_id }}     
+  #  # IMPORTANT security flaw if we don't need 'check-run-condition'
+  #   needs: e2e-x86
+  #   concurrency:
+  #     group: aztec-bench-summary-${{ github.ref_name == 'master' && github.run_id || github.ref_name }}-x86
+  #     cancel-in-progress: true
+  #   steps:
+  #     - name: Checkout
+  #       uses: actions/checkout@v4
+  #       with:
+  #         ref: ${{ github.event.pull_request.head.sha }}
+
+  #     - name: "Assemble benchmark summary from uploaded logs"
+  #         command: ./scripts/ci/assemble_e2e_benchmark_earthly.sh
diff --git a/.github/workflows/protocol-circuits-gate-diff.yml b/.github/workflows/protocol-circuits-gate-diff.yml
index 8126a13193c..7905f0e6a1e 100644
--- a/.github/workflows/protocol-circuits-gate-diff.yml
+++ b/.github/workflows/protocol-circuits-gate-diff.yml
@@ -4,7 +4,7 @@ on:
   push:
     branches:
       - master
-  pull_request:
+  pull_request: {}
 
 jobs:
   compare_protocol_circuits_gates:
diff --git a/.vscode/settings.json b/.vscode/settings.json
index 98b15088b50..0c9b214de1e 100644
--- a/.vscode/settings.json
+++ b/.vscode/settings.json
@@ -160,5 +160,6 @@
     "noir/noir-repo/Cargo.toml",
     "noir/noir-repo/acvm-repo/acvm_js/Cargo.toml",
     "avm-transpiler/Cargo.toml"
-  ]
+  ],
+  "cmake.sourceDirectory": "/mnt/user-data/adam/aztec-packages/barretenberg/cpp"
 }
diff --git a/README.md b/README.md
index 38e03d97604..5abbc1903d8 100644
--- a/README.md
+++ b/README.md
@@ -39,7 +39,7 @@ noirup -v TAG_FROM_THE_FILE
 
 This repository uses CircleCI for continuous integration. Build steps are managed using [`build-system`](https://github.com/AztecProtocol/build-system). Small packages are built and tested as part of a docker build operation, while larger ones and end-to-end tests spin up a large AWS spot instance. Each successful build step creates a new docker image that gets tagged with the package name and commit.
 
-All packages need to be included in the [build manifest](`build_manifest.json`), which declares what paths belong to each package, as well as dependencies between packages. When the CI runs, if none of the rebuild patterns or dependencies were changed, then the build step is skipped and the last successful image is re-tagged with the current commit. Read more on the [`build-system`](https://github.com/AztecProtocol/build-system) repository README.
+All packages need to be included in the [build manifest](`build_manifest.yml`), which declares what paths belong to each package, as well as dependencies between packages. When the CI runs, if none of the rebuild patterns or dependencies were changed, then the build step is skipped and the last successful image is re-tagged with the current commit. Read more on the [`build-system`](https://github.com/AztecProtocol/build-system) repository README.
 
 It is faster to debug CI failures within a persistent ssh session compared to pushing and waiting. You can create a session with "Rerun step with SSH" on CircleCI which will generate an ssh command for debugging on a worker. Run that command locally and then do
 
diff --git a/noir-projects/bootstrap.sh b/noir-projects/bootstrap.sh
index c27610cb8ef..2a46c6e26f2 100755
--- a/noir-projects/bootstrap.sh
+++ b/noir-projects/bootstrap.sh
@@ -48,7 +48,4 @@ else
   for job in $(jobs -p); do
     wait $job || exit 1
   done
-fi
-
-
-
+fi
\ No newline at end of file
diff --git a/scripts/ci/assemble_e2e_benchmark.sh b/scripts/ci/assemble_e2e_benchmark.sh
index 811bef01e6f..6ba69863526 100755
--- a/scripts/ci/assemble_e2e_benchmark.sh
+++ b/scripts/ci/assemble_e2e_benchmark.sh
@@ -10,17 +10,19 @@ BUCKET_NAME="aztec-ci-artifacts"
 LOG_FOLDER="${LOG_FOLDER:-log}"
 BENCH_FOLDER="${BENCH_FOLDER:-bench}"
 COMMIT_HASH="${COMMIT_HASH:-$(git rev-parse HEAD)}"
-BASE_COMMIT_HASH=""
+BASE_BENCH_PATH=""
 BENCHMARK_FILE_JSON="${BENCH_FOLDER}/benchmark.json"
 BASE_BENCHMARK_FILE_JSON="${BENCH_FOLDER}/base-benchmark.json"
 
 # Paths from build-system/scripts/upload_logs_to_s3
 if [ "${CIRCLE_BRANCH:-}" = "master" ]; then
   LOG_SOURCE_FOLDER="logs-v1/master/$COMMIT_HASH"
+  BARRETENBERG_BENCH_SOURCE_FOLDER="barretenberg-bench-v1/master/$COMMIT_HASH"
   BENCHMARK_TARGET_FILE="benchmarks-v1/master/$COMMIT_HASH.json"
   BENCHMARK_LATEST_FILE="benchmarks-v1/latest.json"
 elif [ -n "${CIRCLE_PULL_REQUEST:-}" ]; then
   LOG_SOURCE_FOLDER="logs-v1/pulls/${CIRCLE_PULL_REQUEST##*/}"
+  BARRETENBERG_BENCH_SOURCE_FOLDER="barretenberg-bench-v1/pulls/${CIRCLE_PULL_REQUEST##*/}"
   BENCHMARK_TARGET_FILE="benchmarks-v1/pulls/${CIRCLE_PULL_REQUEST##*/}.json"
 elif [ -n "${CIRCLE_TAG:-}" ]; then
   echo "Skipping benchmark run for ${CIRCLE_TAG} tagged release."
@@ -44,9 +46,12 @@ EXPECTED_LOGS_COUNT=$(find yarn-project/end-to-end/src -type f -name "bench*.tes
 DOWNLOADED_LOGS_COUNT=$(find $LOG_FOLDER -type f -name "*.jsonl" | wc -l)
 if [ "$DOWNLOADED_LOGS_COUNT" -lt "$EXPECTED_LOGS_COUNT" ]; then
   echo Found $DOWNLOADED_LOGS_COUNT out of $EXPECTED_LOGS_COUNT benchmark log files in s3://${BUCKET_NAME}/${LOG_SOURCE_FOLDER}/. Exiting.
-  exit 0
+  exit 1
 fi
 
+# Download barretenberg log files, these are direct benchmarks and separate from the above
+aws s3 cp "s3://${BUCKET_NAME}/${BARRETENBERG_BENCH_SOURCE_FOLDER}/" $LOG_FOLDER --exclude '*' --include '*_bench.json' --recursive
+
 # Generate the aggregated benchmark file
 mkdir -p $BENCH_FOLDER
 CONTAINER_BENCH_FOLDER="/usr/src/yarn-project/bench"
@@ -56,7 +61,7 @@ export DOCKER_RUN_OPTS="\
  -e BENCH_FOLDER=${CONTAINER_BENCH_FOLDER} \
  -v $(realpath $LOG_FOLDER):${CONTAINER_LOG_FOLDER}:rw \
  -e LOG_FOLDER=${CONTAINER_LOG_FOLDER} \
- -e BASE_COMMIT_HASH \
+ -e BASE_BENCH_PATH \
  -e AZTEC_BOT_COMMENTER_GITHUB_TOKEN \
  -e CIRCLE_PULL_REQUEST"
 yarn-project/scripts/run_script.sh workspace @aztec/scripts bench-aggregate
@@ -82,13 +87,13 @@ if [ -n "${CIRCLE_PULL_REQUEST:-}" ]; then
     aws s3 cp "s3://${BUCKET_NAME}/benchmarks-v1/master/$commit_hash.json" $BASE_BENCHMARK_FILE_JSON
     if [ $? -eq 0 ]; then
       echo "Downloaded base data from commit $commit_hash"
-      export BASE_COMMIT_HASH=$commit_hash
+      export BASE_BENCH_PATH=master/$commit_hash
       break;
     fi
   done
   set -e
 
-  if [ -z "${BASE_COMMIT_HASH:-}" ]; then
+  if [ -z "${BASE_BENCH_PATH:-}" ]; then
     echo "No base commit data found"
   fi
 
diff --git a/scripts/ci/upload_logs_to_s3 b/scripts/ci/upload_logs_to_s3
new file mode 100755
index 00000000000..b13c0ebd456
--- /dev/null
+++ b/scripts/ci/upload_logs_to_s3
@@ -0,0 +1,32 @@
+#!/usr/bin/env bash
+
+[ -n "${BUILD_SYSTEM_DEBUG:-}" ] && set -x # conditionally trace
+
+# Uploads to S3 the contents of the log file mounted on the end-to-end container,
+# which contains log entries with an associated event and metrics for it.
+# Logs are uploaded to aztec-ci-artifacts/logs-v1/master/$COMMIT/$JOB.jsonl
+# or to aztec-ci-artifacts/logs-v1/pulls/$PRNUMBER/$JOB.jsonl if on a PR
+
+set -eu
+
+LOG_FOLDER=$1
+BUCKET_NAME="aztec-ci-artifacts"
+COMMIT_HASH="${COMMIT_HASH:-$(git rev-parse HEAD)}"
+
+if [ ! -d "$LOG_FOLDER" ] || [ -z "$(ls -A "$LOG_FOLDER")" ]; then
+  echo "No logs in folder $LOG_FOLDER to upload"
+  exit 0
+fi
+
+# Paths used in scripts/ci/assemble_e2e_benchmark.sh
+if [ "${BRANCH:-}" = "master" ]; then
+  TARGET_FOLDER="logs-v1/master/$COMMIT_HASH/"
+elif [ -n "${PULL_REQUEST:-}" ]; then
+  TARGET_FOLDER="logs-v1/pulls/${PULL_REQUEST##*/}"
+fi
+
+if [ -n "${TARGET_FOLDER:-}" ]; then
+  aws s3 cp $LOG_FOLDER "s3://${BUCKET_NAME}/${TARGET_FOLDER}"  --include "*.jsonl" --recursive
+else
+  echo Skipping upload since no target folder was defined
+fi
\ No newline at end of file
diff --git a/yarn-project/Earthfile b/yarn-project/Earthfile
index e37d7332233..7c883bf9a1a 100644
--- a/yarn-project/Earthfile
+++ b/yarn-project/Earthfile
@@ -32,7 +32,7 @@ build:
     SAVE ARTIFACT /usr/src
 
 # TODO versioning flow at end before publish?
-# ENV COMMIT_TAG=$EARTHLY_BUILD_SHA
+# ENV COMMIT_TAG=$EARTHLY_GIT_HASH
 # RUN ./scripts/version_packages.sh
 
 # run:
@@ -86,4 +86,4 @@ build-end-to-end:
     FROM +end-to-end-minimal
     SAVE IMAGE --push aztecprotocol/end-to-end-cache:$EARTHLY_GIT_HASH
     FROM +aztec
-    SAVE IMAGE --push aztecprotocol/aztec-cache:$EARTHLY_GIT_HASH
\ No newline at end of file
+    SAVE IMAGE --push aztecprotocol/aztec-cache:$EARTHLY_GIT_HASH
diff --git a/yarn-project/circuit-types/src/stats/metrics.ts b/yarn-project/circuit-types/src/stats/metrics.ts
index 4ee8ea7e04c..467c80c7839 100644
--- a/yarn-project/circuit-types/src/stats/metrics.ts
+++ b/yarn-project/circuit-types/src/stats/metrics.ts
@@ -8,7 +8,8 @@ export type MetricGroupBy =
   | 'classes-registered'
   | 'leaf-count'
   | 'data-writes'
-  | 'fee-payment-method';
+  | 'fee-payment-method'
+  | 'circuit-size-in-gates';
 
 /** Definition of a metric to track in benchmarks. */
 export interface Metric {
@@ -24,6 +25,18 @@ export interface Metric {
 
 /** Metric definitions to track from benchmarks. */
 export const Metrics = [
+  {
+    name: 'client_ivc_proving_time_in_ms',
+    groupBy: 'circuit-size-in-gates',
+    description: 'Proving time for ClientIVC grouped by circuit size.',
+    events: [],
+  },
+  {
+    name: 'ultrahonk_proving_time_in_ms',
+    groupBy: 'circuit-size-in-gates',
+    description: 'Proving time for UltraHonk grouped by circuit size.',
+    events: [],
+  },
   {
     name: 'l1_rollup_calldata_size_in_bytes',
     groupBy: 'block-size',
diff --git a/yarn-project/end-to-end/Earthfile b/yarn-project/end-to-end/Earthfile
index e5ad686c60e..d646e2e9465 100644
--- a/yarn-project/end-to-end/Earthfile
+++ b/yarn-project/end-to-end/Earthfile
@@ -259,11 +259,11 @@ bench-publish-rollup:
   DO +E2E_TEST --test=benchmarks/bench_publish_rollup.test.ts --debug="aztec:benchmarks:*,aztec:sequencer,aztec:sequencer:*,aztec:world_state,aztec:merkle_trees" --e2e_mode=$e2e_mode  --compose_file=./scripts/docker-compose-no-sandbox.yml 
 
 # TODO need to investigate why this isn't working
-# bench-process-history:
-#   ARG e2e_mode=local
-#   DO +E2E_TEST --test=benchmarks/bench_process_history.test.ts --debug="aztec:benchmarks:*,aztec:sequencer,aztec:sequencer:*,aztec:world_state,aztec:merkle_trees" --e2e_mode=$e2e_mode  --compose_file=./scripts/docker-compose-no-sandbox.yml 
+bench-process-history:
+  ARG e2e_mode=local
+  DO +E2E_TEST --test=benchmarks/bench_process_history.test.ts --debug="aztec:benchmarks:*,aztec:sequencer,aztec:sequencer:*,aztec:world_state,aztec:merkle_trees" --e2e_mode=$e2e_mode  --compose_file=./scripts/docker-compose-no-sandbox.yml 
 
 # TODO need to investigate why this isn't working
-# bench-tx-size:
-#   ARG e2e_mode=local
-#   DO +E2E_TEST --test=benchmarks/bench_tx_size_fees.test.ts --debug="aztec:benchmarks:*,aztec:sequencer,aztec:sequencer:*,aztec:world_state,aztec:merkle_trees" --e2e_mode=$e2e_mode  --compose_file=./scripts/docker-compose-no-sandbox.yml 
+bench-tx-size:
+  ARG e2e_mode=local
+  DO +E2E_TEST --test=benchmarks/bench_tx_size_fees.test.ts --debug="aztec:benchmarks:*,aztec:sequencer,aztec:sequencer:*,aztec:world_state,aztec:merkle_trees" --e2e_mode=$e2e_mode  --compose_file=./scripts/docker-compose-no-sandbox.yml 
diff --git a/yarn-project/end-to-end/scripts/docker-compose.yml b/yarn-project/end-to-end/scripts/docker-compose.yml
index 054e16477f8..cc33a29709f 100644
--- a/yarn-project/end-to-end/scripts/docker-compose.yml
+++ b/yarn-project/end-to-end/scripts/docker-compose.yml
@@ -44,7 +44,6 @@ services:
         yarn test ${TEST:-./src/e2e_deploy_contract.test.ts}
       '
     volumes:
-      # TODO(AD) currently earthly uses /build instead of /usr/src
       - ../log:/usr/src/yarn-project/end-to-end/log:rw
     depends_on:
       - sandbox
diff --git a/yarn-project/scripts/src/benchmarks/aggregate.ts b/yarn-project/scripts/src/benchmarks/aggregate.ts
index ce1d8905463..6440cb14e36 100644
--- a/yarn-project/scripts/src/benchmarks/aggregate.ts
+++ b/yarn-project/scripts/src/benchmarks/aggregate.ts
@@ -228,6 +228,15 @@ function processEntry(entry: Stats, results: BenchmarkCollectedResults, fileName
   }
 }
 
+function getBarretenbergMetric(context: { executable?: string }): MetricName | undefined {
+  if (context.executable?.includes('ultra_honk')) {
+    return 'ultrahonk_proving_time_in_ms';
+  } else if (context.executable?.includes('client_ivc')) {
+    return 'client_ivc_proving_time_in_ms';
+  }
+  return undefined;
+}
+
 /** Array of collected raw results for a given metric. */
 type BenchmarkCollectedMetricResults = Record<string, number[]>;
 
@@ -253,7 +262,8 @@ export async function main() {
     }
   }
 
-  log(`Collected entries: ${JSON.stringify(collected)}`);
+  // Spammy if on by default
+  // log(`Collected entries: ${JSON.stringify(collected)}`);
 
   // For each bucket of each metric compute the average all collected data points
   const results: BenchmarkResults = {};
@@ -269,6 +279,28 @@ export async function main() {
     }
   }
 
+  // Add google benchmark json files, which have data already averaged
+  const googleBenchmarkFiles = fs.readdirSync(LogsDir).filter(f => f.endsWith('.json'));
+  for (const file of googleBenchmarkFiles) {
+    const data = JSON.parse(fs.readFileSync(path.join(LogsDir, file), 'utf-8'));
+    if (!data.context || !data.benchmarks) {
+      log(`Invalid google benchmark file: ${file}`);
+      continue;
+    }
+    const metric = getBarretenbergMetric(data.context);
+    if (!metric) {
+      log(`Unknown executable in benchmark file ${file}: ${data.context.executable}`);
+      continue;
+    }
+    const circuitSize = '2^20'; // Where to load size from?
+    const value = data.benchmarks[0]?.real_time;
+    if (value === undefined) {
+      log(`Couldn't find real_time in benchmark file ${file}`);
+      continue;
+    }
+    results[metric] = { [circuitSize]: value };
+  }
+
   const timestampedResults: BenchmarkResultsWithTimestamp = { ...results, timestamp: new Date().toISOString() };
 
   // Write results to disk
diff --git a/yarn-project/scripts/src/benchmarks/markdown.ts b/yarn-project/scripts/src/benchmarks/markdown.ts
index 2887805cfdc..f803cd683ca 100644
--- a/yarn-project/scripts/src/benchmarks/markdown.ts
+++ b/yarn-project/scripts/src/benchmarks/markdown.ts
@@ -189,12 +189,13 @@ export function getMarkdown() {
   const metricsByClassesRegistered = Metrics.filter(m => m.groupBy === 'classes-registered').map(m => m.name);
   const metricsByFeePaymentMethod = Metrics.filter(m => m.groupBy === 'fee-payment-method').map(m => m.name);
   const metricsByLeafCount = Metrics.filter(m => m.groupBy === 'leaf-count').map(m => m.name);
+  const metricsByCircuitSize = Metrics.filter(m => m.groupBy === 'circuit-size-in-gates').map(m => m.name);
 
   const metricsTxPxeProcessing = Metrics.filter(m => m.name === 'tx_pxe_processing_time_ms').map(m => m.name);
   const metricsTxSeqProcessing = Metrics.filter(m => m.name === 'tx_sequencer_processing_time_ms').map(m => m.name);
 
-  const baseHash = process.env.BASE_COMMIT_HASH;
-  const baseUrl = baseHash && `[\`${baseHash.slice(0, 8)}\`](${S3_URL}/benchmarks-v1/master/${baseHash}.json)`;
+  const baseUrlPath = process.env.BASE_BENCH_PATH;
+  const baseUrl = baseUrlPath && `[\`${baseUrlPath.slice(0, 8)}\`](${S3_URL}/benchmarks-v1/master/${baseUrlPath}.json)`;
   const baseCommitText = baseUrl
     ? `\nValues are compared against data from master at commit ${baseUrl} and shown if the difference exceeds 1%.`
     : '';
@@ -214,10 +215,15 @@ ${getWarningsSummary(benchmark, baseBenchmark)}
 
 <summary>Detailed results</summary>
 
-All benchmarks are run on txs on the \`Benchmarking\` contract on the repository. Each tx consists of a batch call  to \`create_note\` and \`increment_balance\`, which guarantees that each tx has a private call, a nested private call, a public call, and a nested public call, as well as an emitted private note, an unencrypted log, and public storage read and write.
+Except for proving times, all benchmarks are run on txs on the \`Benchmarking\` contract on the repository. Each tx consists of a batch call  to \`create_note\` and \`increment_balance\`, which guarantees that each tx has a private call, a nested private call, a public call, and a nested public call, as well as an emitted private note, an unencrypted log, and public storage read and write.
 ${prSourceDataText}
 ${baseCommitText}
 
+### Proving times
+
+Average proving times for benchmarks measured in bb.
+${getTableContent(pick(benchmark, metricsByCircuitSize), baseBenchmark, 'gates', 'Circuit')}
+
 ### L2 block published to L1
 
 Each column represents the number of txs on an L2 block published to L1.